added 16-bit sRGB/XYZ conversion
need to get it working on images though
This commit is contained in:
John Cupitt
parent
e0b3b4157f
commit
cb4965536d
@ -12,6 +12,8 @@
|
||||
* - sRGB only, support for other RGBs is now via lcms
|
||||
* 1/11/12
|
||||
* - faster and more accurate sRGB <-> XYZ conversion
|
||||
* 6/11/12
|
||||
* - added a 16-bit path
|
||||
*/
|
||||
|
||||
/*
|
||||
@ -72,26 +74,34 @@ static float vips_v2Y_8[256];
|
||||
*
|
||||
* There's an extra element at the end to let us do a +1 for interpolation.
|
||||
*/
|
||||
static int vips_16_Y2v[65536 + 1];
|
||||
static int vips_Y2v_16[65536 + 1];
|
||||
|
||||
/* 16-bit sRGB -> linear lut.
|
||||
*/
|
||||
static float vips_16_v2Y[65536];
|
||||
static float vips_v2Y_16[65536];
|
||||
|
||||
/* linear RGB -> XYZ matrix.
|
||||
/* We need Y in 0 - 100. We can save three muls later if we pre-scale the
|
||||
* matrix.
|
||||
*
|
||||
* The matrix already includes the D65 channel weighting, so we just scale by
|
||||
* Y.
|
||||
*/
|
||||
#define SCALE (VIPS_D65_Y0 / 3.0)
|
||||
|
||||
/* linear RGB -> XYZ matrix.
|
||||
*/
|
||||
static float vips_mat_RGB2XYZ[3][3] = {
|
||||
{ 0.4124, 0.3576, 0.18056 },
|
||||
{ 0.2126, 0.7152, 0.0722 },
|
||||
{ 0.0193, 0.1192, 0.9505 }
|
||||
{ SCALE * 0.4124, SCALE * 0.3576, SCALE * 0.18056 },
|
||||
{ SCALE * 0.2126, SCALE * 0.7152, SCALE * 0.0722 },
|
||||
{ SCALE * 0.0193, SCALE * 0.1192, SCALE * 0.9505 }
|
||||
};
|
||||
|
||||
/* XYZ -> linear RGB matrix.
|
||||
*/
|
||||
static float vips_mat_XYZ2RGB[3][3] = {
|
||||
{ 3.2406, -1.5372, -0.4986 },
|
||||
{ -0.9689, 1.8758, 0.0415, },
|
||||
{ 0.0557, -0.2040, 1.0570 }
|
||||
{ 3.2406 / SCALE, -1.5372 / SCALE, -0.4986 / SCALE },
|
||||
{ -0.9689 / SCALE, 1.8758 / SCALE, 0.0415 / SCALE },
|
||||
{ 0.0557 / SCALE, -0.2040 / SCALE, 1.0570 / SCALE }
|
||||
};
|
||||
|
||||
/* Do our own indexing of the arrays below to make sure we get efficient mults.
|
||||
@ -104,15 +114,15 @@ static VipsPel vips_red[64 * 64 * 64];
|
||||
static VipsPel vips_green[64 * 64 * 64];
|
||||
static VipsPel vips_blue[64 * 64 * 64];
|
||||
|
||||
/* Create the sRGB linear and unlinear luts.
|
||||
/* Create the sRGB linear and unlinear luts. @range is eg. 256 for 8-bit luts.
|
||||
*/
|
||||
static void *
|
||||
calcul_tables_8( void *client )
|
||||
static void
|
||||
calcul_tables( int range, int *Y2v, float *v2Y )
|
||||
{
|
||||
int i;
|
||||
|
||||
for( i = 0; i < 256; i++ ) {
|
||||
float f = (float) i / (256 - 1);
|
||||
for( i = 0; i < range; i++ ) {
|
||||
float f = (float) i / (range - 1);
|
||||
float v;
|
||||
|
||||
if( f <= 0.0031308 )
|
||||
@ -120,22 +130,28 @@ calcul_tables_8( void *client )
|
||||
else
|
||||
v = (1.0 + 0.055) * pow( f, 1.0 / 2.4 ) - 0.055;
|
||||
|
||||
vips_Y2v_8[i] = 255.0 * v;
|
||||
Y2v[i] = (range - 1) * v;
|
||||
}
|
||||
|
||||
/* Copy the final element. This is used in the piecewise linear
|
||||
* interpolator below.
|
||||
*/
|
||||
vips_Y2v_8[256] = vips_Y2v_8[256 - 1];
|
||||
Y2v[range] = Y2v[range - 1];
|
||||
|
||||
for( i = 0; i < 256; i++ ) {
|
||||
float f = i / 255.0;
|
||||
for( i = 0; i < range; i++ ) {
|
||||
float f = (float) i / (range - 1);
|
||||
|
||||
if( f <= 0.04045 )
|
||||
vips_v2Y_8[i] = f / 12.92;
|
||||
v2Y[i] = f / 12.92;
|
||||
else
|
||||
vips_v2Y_8[i] = pow( (f + 0.055) / (1 + 0.055), 2.4 );
|
||||
v2Y[i] = pow( (f + 0.055) / (1 + 0.055), 2.4 );
|
||||
}
|
||||
}
|
||||
|
||||
static void *
|
||||
calcul_tables_8( void *client )
|
||||
{
|
||||
calcul_tables( 256, vips_Y2v_8, vips_v2Y_8 );
|
||||
|
||||
return( NULL );
|
||||
}
|
||||
@ -158,28 +174,26 @@ vips_col_make_tables_RGB_8( void )
|
||||
|
||||
/* Computes the transform: r,g,b => Yr,Yg,Yb. It finds Y values in
|
||||
* lookup tables and calculates X, Y, Z.
|
||||
*
|
||||
* @range is eg. 256 for 8-bit data,
|
||||
*/
|
||||
int
|
||||
vips_col_sRGB2XYZ_8( int r, int g, int b, float *X, float *Y, float *Z )
|
||||
static int
|
||||
vips_col_sRGB2XYZ( int range, float *lut,
|
||||
int r, int g, int b, float *X, float *Y, float *Z )
|
||||
{
|
||||
float *lut = vips_v2Y_8;
|
||||
|
||||
int maxval = range - 1;
|
||||
float Yr, Yg, Yb;
|
||||
int i;
|
||||
|
||||
vips_col_make_tables_RGB_8();
|
||||
|
||||
i = VIPS_CLIP( 0, r, 255 );
|
||||
i = VIPS_CLIP( 0, r, maxval );
|
||||
Yr = lut[i];
|
||||
|
||||
i = VIPS_CLIP( 0, g, 255 );
|
||||
i = VIPS_CLIP( 0, g, maxval );
|
||||
Yg = lut[i];
|
||||
|
||||
i = VIPS_CLIP( 0, b, 255 );
|
||||
i = VIPS_CLIP( 0, b, maxval );
|
||||
Yb = lut[i];
|
||||
|
||||
/* The matrix already includes D65 channel weighting.
|
||||
*/
|
||||
*X = vips_mat_RGB2XYZ[0][0] * Yr +
|
||||
vips_mat_RGB2XYZ[0][1] * Yg + vips_mat_RGB2XYZ[0][2] * Yb;
|
||||
*Y = vips_mat_RGB2XYZ[1][0] * Yr +
|
||||
@ -187,39 +201,75 @@ vips_col_sRGB2XYZ_8( int r, int g, int b, float *X, float *Y, float *Z )
|
||||
*Z = vips_mat_RGB2XYZ[2][0] * Yr +
|
||||
vips_mat_RGB2XYZ[2][1] * Yg + vips_mat_RGB2XYZ[2][2] * Yb;
|
||||
|
||||
*X *= VIPS_D65_Y0;
|
||||
*Y *= VIPS_D65_Y0;
|
||||
*Z *= VIPS_D65_Y0;
|
||||
|
||||
return( 0 );
|
||||
}
|
||||
|
||||
/* Computes the transform: r,g,b => Yr,Yg,Yb. It finds Y values in
|
||||
* lookup tables and calculates X, Y, Z.
|
||||
*
|
||||
* rgb are in the range 0 to 255,
|
||||
*/
|
||||
int
|
||||
vips_col_sRGB2XYZ_8( int r, int g, int b, float *X, float *Y, float *Z )
|
||||
{
|
||||
vips_col_make_tables_RGB_8();
|
||||
|
||||
return( vips_col_sRGB2XYZ( 256, vips_v2Y_8, r, g, b, X, Y, Z ) );
|
||||
}
|
||||
|
||||
static void *
|
||||
calcul_tables_16( void *client )
|
||||
{
|
||||
calcul_tables( 65536, vips_Y2v_16, vips_v2Y_16 );
|
||||
|
||||
return( NULL );
|
||||
}
|
||||
|
||||
static void
|
||||
vips_col_make_tables_RGB_16( void )
|
||||
{
|
||||
static gboolean made_tables = FALSE;
|
||||
|
||||
/* We want to avoid having a mutex in this path, so use gonce and a
|
||||
* static var instead.
|
||||
*/
|
||||
if( !made_tables ) {
|
||||
static GOnce once = G_ONCE_INIT;
|
||||
|
||||
(void) g_once( &once, calcul_tables_16, NULL );
|
||||
made_tables = TRUE;
|
||||
}
|
||||
}
|
||||
|
||||
/* Computes the transform: r,g,b => Yr,Yg,Yb. It finds Y values in
|
||||
* lookup tables and calculates X, Y, Z.
|
||||
*
|
||||
* rgb are in the range 0 to 65535,
|
||||
*/
|
||||
int
|
||||
vips_col_sRGB2XYZ_16( int r, int g, int b, float *X, float *Y, float *Z )
|
||||
{
|
||||
vips_col_make_tables_RGB_16();
|
||||
|
||||
return( vips_col_sRGB2XYZ( 65536, vips_v2Y_16, r, g, b, X, Y, Z ) );
|
||||
}
|
||||
|
||||
/* Turn XYZ into display colour. Return or=1 for out of gamut - rgb will
|
||||
* contain an approximation of the right colour.
|
||||
*/
|
||||
int
|
||||
vips_col_XYZ2sRGB_8( float X, float Y, float Z,
|
||||
int *r_ret, int *g_ret, int *b_ret,
|
||||
static int
|
||||
vips_col_XYZ2sRGB( int range, int *lut,
|
||||
float X, float Y, float Z,
|
||||
int *r, int *g, int *b,
|
||||
int *or_ret )
|
||||
{
|
||||
int *lut = vips_Y2v_8;
|
||||
|
||||
int or;
|
||||
int maxval = range - 1;
|
||||
|
||||
float Yr, Yg, Yb;
|
||||
int or;
|
||||
float Yf;
|
||||
int Yi;
|
||||
float v;
|
||||
int r, g, b;
|
||||
|
||||
vips_col_make_tables_RGB_8();
|
||||
|
||||
/* The matrix already includes D65 channel weighting, just change from
|
||||
* 0 - 100 to 0 - 1.
|
||||
*/
|
||||
X = X / VIPS_D65_Y0;
|
||||
Y = Y / VIPS_D65_Y0;
|
||||
Z = Z / VIPS_D65_Y0;
|
||||
|
||||
/* Multiply through the matrix to get luminosity values.
|
||||
*/
|
||||
@ -251,33 +301,61 @@ vips_col_XYZ2sRGB_8( float X, float Y, float Z,
|
||||
|
||||
or = 0;
|
||||
|
||||
Yf = Yr * (256 - 1);
|
||||
CLIP( 0, Yf, 256 - 1);
|
||||
Yf = Yr * maxval;
|
||||
CLIP( 0, Yf, maxval);
|
||||
Yi = (int) Yf;
|
||||
v = lut[Yi] + (lut[Yi + 1] - lut[Yi]) * (Yf - Yi);
|
||||
r = VIPS_RINT( v );
|
||||
*r = VIPS_RINT( v );
|
||||
|
||||
Yf = Yg * (256 - 1);
|
||||
CLIP( 0, Yf, 256 - 1);
|
||||
Yf = Yg * maxval;
|
||||
CLIP( 0, Yf, maxval);
|
||||
Yi = (int) Yf;
|
||||
v = lut[Yi] + (lut[Yi + 1] - lut[Yi]) * (Yf - Yi);
|
||||
g = VIPS_RINT( v );
|
||||
*g = VIPS_RINT( v );
|
||||
|
||||
Yf = Yb * (256 - 1);
|
||||
CLIP( 0, Yf, 256 - 1);
|
||||
Yf = Yb * maxval;
|
||||
CLIP( 0, Yf, maxval);
|
||||
Yi = (int) Yf;
|
||||
v = lut[Yi] + (lut[Yi + 1] - lut[Yi]) * (Yf - Yi);
|
||||
b = VIPS_RINT( v );
|
||||
*b = VIPS_RINT( v );
|
||||
|
||||
*r_ret = r;
|
||||
*g_ret = g;
|
||||
*b_ret = b;
|
||||
|
||||
*or_ret = or;
|
||||
if( or_ret )
|
||||
*or_ret = or;
|
||||
|
||||
return( 0 );
|
||||
}
|
||||
|
||||
/* Turn XYZ into display colour. Return or=1 for out of gamut - rgb will
|
||||
* contain an approximation of the right colour.
|
||||
*
|
||||
* r, g, b are scaled to fit the range 0 - 255.
|
||||
*/
|
||||
int
|
||||
vips_col_XYZ2sRGB_8( float X, float Y, float Z,
|
||||
int *r, int *g, int *b,
|
||||
int *or )
|
||||
{
|
||||
vips_col_make_tables_RGB_8();
|
||||
|
||||
return( vips_col_XYZ2sRGB( 256, vips_Y2v_8, X, Y, Z, r, g, b, or ) );
|
||||
}
|
||||
|
||||
/* Turn XYZ into display colour. Return or=1 for out of gamut - rgb will
|
||||
* contain an approximation of the right colour.
|
||||
*
|
||||
* r, g, b are scaled to fit the range 0 - 65535.
|
||||
*/
|
||||
int
|
||||
vips_col_XYZ2sRGB_16( float X, float Y, float Z,
|
||||
int *r, int *g, int *b,
|
||||
int *or )
|
||||
{
|
||||
vips_col_make_tables_RGB_16();
|
||||
|
||||
return( vips_col_XYZ2sRGB( 65536, vips_Y2v_16,
|
||||
X, Y, Z, r, g, b, or ) );
|
||||
}
|
||||
|
||||
/* Build Lab->disp tables.
|
||||
*/
|
||||
static void *
|
||||
|
||||
@ -54,10 +54,11 @@ typedef VipsColourCodeClass VipssRGB2XYZClass;
|
||||
|
||||
G_DEFINE_TYPE( VipssRGB2XYZ, vips_sRGB2XYZ, VIPS_TYPE_COLOUR_CODE );
|
||||
|
||||
/* Convert a buffer.
|
||||
/* Convert a buffer of 8-bit pixels.
|
||||
*/
|
||||
static void
|
||||
vips_sRGB2XYZ_line( VipsColour *colour, VipsPel *out, VipsPel **in, int width )
|
||||
vips_sRGB2XYZ_line_8( VipsColour *colour,
|
||||
VipsPel *out, VipsPel **in, int width )
|
||||
{
|
||||
VipsPel *p = in[0];
|
||||
float *q = (float *) out;
|
||||
@ -82,6 +83,35 @@ vips_sRGB2XYZ_line( VipsColour *colour, VipsPel *out, VipsPel **in, int width )
|
||||
}
|
||||
}
|
||||
|
||||
/* Convert a buffer of 16-bit pixels.
|
||||
*/
|
||||
static void
|
||||
vips_sRGB2XYZ_line_16( VipsColour *colour,
|
||||
VipsPel *out, VipsPel **in, int width )
|
||||
{
|
||||
unsigned short *p = in[0];
|
||||
float *q = (float *) out;
|
||||
|
||||
int i;
|
||||
|
||||
for( i = 0; i < width; i++ ) {
|
||||
int r = p[0];
|
||||
int g = p[1];
|
||||
int b = p[2];
|
||||
float X, Y, Z;
|
||||
|
||||
p += 3;
|
||||
|
||||
vips_col_sRGB2XYZ_16( r, g, b, &X, &Y, &Z );
|
||||
|
||||
q[0] = X;
|
||||
q[1] = Y;
|
||||
q[2] = Z;
|
||||
|
||||
q += 3;
|
||||
}
|
||||
}
|
||||
|
||||
static void
|
||||
vips_sRGB2XYZ_class_init( VipssRGB2XYZClass *class )
|
||||
{
|
||||
|
||||
@ -194,10 +194,13 @@ float vips_col_Ccmc2C( float Ccmc );
|
||||
float vips_col_Chcmc2h( float C, float hcmc );
|
||||
|
||||
int vips_col_XYZ2sRGB_8( float X, float Y, float Z,
|
||||
int *r_ret, int *g_ret, int *b_ret,
|
||||
int *or_ret );
|
||||
int *r, int *g, int *b, int *or_ret );
|
||||
int vips_col_sRGB2XYZ_8( int r, int g, int b, float *X, float *Y, float *Z );
|
||||
|
||||
int vips_col_XYZ2sRGB_16( float X, float Y, float Z,
|
||||
int *r, int *g, int *b, int *or_ret );
|
||||
int vips_col_sRGB2XYZ_16( int r, int g, int b, float *X, float *Y, float *Z );
|
||||
|
||||
float vips_pythagoras( float L1, float a1, float b1,
|
||||
float L2, float a2, float b2 );
|
||||
float vips_col_dE00(
|
||||
|
||||
Loading…
Reference in New Issue
Block a user